Materials and Methods

The population studied is derived from the cross of Probus, a mutant carrying a male sterile allele (ms1b), with 59 lines covering a large genetic diversity. Resulting F1 progenies were alternatively selfed (F2) or back-crossed (BC1) with the 59 parents to reduce the Probus genome contribution to the population. These two progenies (F2 & BC1) were sown together in an isolated field surrounded by rye. Male sterile spikes were tagged during flowering and harvested at maturity. Then, and over 12 generations, a random sample of harvested seeds on male sterile plants was drawn each year, and resown in fall, in order to reach between 5,000 and 10,000 adult plants (2,500–5,000 male-sterile plants). After 12 generations we derived 1,000 SSD lines (F5, Fig. 38.1).

Fig. 38.1 Population creation scheme

First we inferred allelic frequencies in the initial population on the basis of the 56 parental lines (including Probus) still available in seed banks (four missing ones), estimating their contributions to the global pool using a Bayesian method (Thépot et al. 2015). The evolved population was studied through a subset of 380 SSD lines, representative of the phenotypic diversity of the 1,000 lines.

Vernalization requirement was assessed in field trial at Le Moulon over two seasons (2010–2011 and 2011–2012), with a spring sowing (April), on a single row of 20 seeds per genotype. For each row, the heading date was scored when half of the plants had half of the main ear emerged from the flag leaf. The heading date was transformed into sums of degree-days (dd) (sums of the mean temperature per day) from sowing to heading. On the basis of the bimodal distribution of the heading date (Fig. 38.2), SSD lines were classified as spring type (heading before 2,000dd), or winter type for the others. Genotypes with inconsistent behavior between both years were discarded (eight SSD lines).

Genotyping was performed using the 9 K i-select SNP array. Only SNPs unambiguously scored as biallelic after a visual inspection using Genome Studio software,

Fig. 38.2 Heading date distribution of all genotypes for spring sowing in 2011, the condition with the least vernalization to determine winter and spring type

were kept. Using KASPar SNP genotyping system (KBioscience), 14 additional polymorphisms located in candidate genes (earliness pathway such as PPD or VRN families) were genotyped.

The diversity detected by SNPs in both populations (parental lines and SSD lines) was compared using Minor Allele Frequency (MAF) and expected heterozygosity (He, Nei diversity). Evolution of growth habit was tested through a comparison of spring/winter ratio in the initial and the evolved populations (Chi square test). Strong shifts in allelic frequencies were used to detect markers under selection, using a new method (Thépot et al. 2015). Q-values were estimated to cope with the multiple tests (Storey and Tibshirani 2003). Each marker under selection was also tested for association with growth habit using a Logit model.

Results and Discussion

The genotyping of 436 lines (56 parents + 380 F5 lines) with the 9 K i-select SNP assay resulted in 7,270 SNPs with high scoring quality. Among these SNPs, 88.4 % were polymorphic in the initial population and 85.8 % in the evolved population. This slightly higher diversity in the initial population was also observed on allelic frequencies of polymorphic SNPs (mean MAF: 0.18 vs. 0.17 and He: 0.25 vs. 0.24). The MAF distribution (Fig. 38.3) showed a globally high frequency of SNPs with a MAF inferior to 0.05, rare alleles being more frequent in the evolved population. This distribution contrasts with the one observed on a worldwide panel, using the same SNP array, which demonstrates a deficit of SNPs with a low MAF (Cavanagh et al. 2013). This deficit might be due to the fact that lines were chosen to maximize

Fig. 38.3 Distribution of minor allele frequency in the initial and evolved populations

the genetic diversity, and SNPs were intentionally selected to favor common allele in a panel of 26 cultivars from mainly USA and Australia (Cavanagh et al. 2013).

Among the 6,476 polymorphic markers, 57 were detected under selection, representing 26 independent genomic areas. When assessing phenotypic evolution for flowering time, we observed a significant shift from 20 % of spring type in the initial population to 47 % in the evolved population. Among markers under selection, three were associated with the growth habit (p-value <0.05). These markers, located on the 5D and 4A chromosomes, only explained a rather limited part of the phenotypic variation (2.8 % with a global model included the three markers). Yet, they all have, experienced an increase of the spring allele frequency in the evolved population which explained a raise of spring type ranging from 3.5 % to 5.4 %. The 5D markers are located on the same chromosome as Vrn-D1 but did not present linkage disequilibrium with the marker located in this gene (r2 <0.009). Surprisingly polymorphisms in candidate genes like Vrn families (Vrn-A1-Prom, Vrn-A1-ex7 and Vrn-D1), although strongly associated to the growth habit (p-value <10−10, r2 = 28 %) have not been detected as targeted by selection. To take into account the complexity of interaction between these three markers, we assumed that spring alleles are both dominant and epistatic (Rousset et al. 2011). Thus as soon as there is at least one spring allele, the haplotype was classified as spring type or winter type otherwise. For parental lines, these Vrn haplotypes are almost completely explaining phenotype (97.5 % of correspondence, r2 = 0.85). However for the evolved population 30

% of SSD lines with winter Vrn haplotype exhibited a spring phenotype (Table 38.1) (r2 = 0.3). This evolution might be explained by (i) high level of recombination that broke the initial full linkage disequilibrium between causal mutations and the three SNPs genotyped, and/or (ii) the increase of spring alleles at one (or several) non-genotyped Vrn genes, such as Vrn-B1. As SNPs, Vrn haplotypes did not present a significant shift (p-value = 0.29), although a 6 % increase of spring haplotypes was observed between the initial and the evolved population (Table 38.1). One hypothesis to explain this absence of significant shift at these candidate genes could be their strong effect: a little variation in frequency at these major genes may have a strong effect on growth habit.

Association genetics and evolutionary approach provided complementary results. The first method detected QTLs with major effects while the second detected QTLs with lower effect but contributing to the evolution of phenotypes. Joint study of phenotypic and genetic evolutions allowed to detect new markers involved in the control of the growth habit on the 5D and 4A chromosomes.

With its high diversity, absence of structure and low LD (Thépot et al. 2015), this population appears as a new QTL mapping resource, allowing the discovery of original genomic regions controlling traits of interest. Ongoing studies will better explore the potential of this population for detection, using the 1,000 SSD lines.

Table 38.1 Summary results of Vrn haplotypes evolution between the initial population and the evolved population and their association to the winter/spring phenotype

Genotypes

Haplotype type

Freq. initial population

Freq. SSD lines

Vrn-A1

promoter

Vrn-A1

exon7

Vrn-D1

S

S

W

Spring

19.6 %

25.3 %

S

H

W

(94 % S; 6 % W)

(93 % S; 7 % W)

S

H

NA

W

S

S

W

S

W

W

S

NA

W

H

S

W

W

S

NA

S

W

W

W

W

Winter

79.3 %

69.7 %

(0.8 % S; 99.2 % W)

(30 % S; 70 % W)

W

W

NA

NA

1.1 %

5 %

W

NA

W

W

H

W

Genotypes are coded with S, H, W, with S for the homozygote spring allele, H for heterozygote and W for homozygote winter allele, assuming that spring allele are dominant and epistatic

 
< Prev   CONTENTS   Next >